1 Reply Latest reply on Jul 29, 2011 10:24 AM by Mircea Markus

    Setting up an Async Replication Cluster for nodes with more than 5GB of data

    fealves78 Newbie

      Hello everyone,

       

      I am a beginner with Infinispan and I would like to get some help on setting up my cache-cluster.

       

      I have the following configuration:

       

      VMLab1.png

       

      * My current cluster has the following node composition (This is a test scenario. Please refer to the image above for more details):

      - COS : 2 Nodes (N1 - 192.168.2.2 and N2 - 192.168.2.3)

      - DEN: 1 Node (N3 - 192.168.3.2)

      - UTAH: 1 Node (N4 - 192.168.4.2)

      - Infinispan 4.2.1

      - Java JDK 1.6.0.26

       

       

      * These nodes need to be replicated and there is a possibility that more nodes and regions will be added to the cluster in the near future. Also, each node needs to hold about 6GB of data (Is that possible? How java app memory / heap size limits interfere with this?)

       

      * I have performed some initial tests with both DISTRIBUTION and REPLICATION modes. The tests work fine for smaller amounts of data, but if it goes over 1.5GB of data per node, everything becomes really slow. Thus, I would like to get some feedback on my configuration, especially for running the cluster in Replication Mode with nodes holding more than 5GB of data. Should a hybrid configuration, such as DISTRIBUTION internally for the regions and REPLICATION between regions, be considered? If so, could you please provide a template configuration for it?

       

      * Also, I would like to know if there are any fine tuning mechanisms for throttling of the connections between nodes (I believe this may be done through JBOSS but i am not sure about it).

       

       

      Attached to this post are the configuration files I am currently using for the 4 nodes.

       

      Thanks for the help!

       

      Francisco.